Skip to main content

The Sample Dataset

This chapter gives a detailed description of the processing of the MNE sample dataset, which can be employed to familiarize with the Workflow.

important

Going through this analysis exercise is not a substitute for reading other chapters of this manual and understanding the concepts underlying MNE software.

Overview​

The MNE software is accompanied by a sample dataset which includes:

  • MRI reconstructions — created with FreeSurfer
  • MEG/EEG data — acquired with the Neuromag Vectorview system at the MGH/HMS/MIT Athinoula A. Martinos Center for Biomedical Imaging

EEG data from a 60-channel electrode cap was acquired simultaneously with MEG. The original MRI dataset was acquired with a Siemens 1.5 T Sonata scanner using an MPRAGE sequence.

Experimental Paradigm​

In the MEG/EEG experiment, checkerboard patterns were presented into the left and right visual field, interspersed by tones to the left or right ear. The interval between stimuli was 750 ms. Occasionally a smiley face was presented at the center of the visual field. The subject was asked to press a key with the right index finger as soon as possible after the appearance of the face.

Trigger Codes​

NameCodeContents
LA1Response to left-ear auditory stimulus
RA2Response to right-ear auditory stimulus
LV3Response to left visual field stimulus
RV4Response to right visual field stimulus
Smiley5Response to the smiley face
Button32Response triggered by the button press

Dataset Contents​

MEG/EEG Data (MEG/sample/)​

FileContents
sample_audvis_raw.fifThe raw MEG/EEG data
audvis.aveTemplate script for off-line averaging
audvis.covTemplate script for noise-covariance matrix computation

MRI Data (subjects/sample/)​

File / DirectoryContents
bem/Directory for the forward modelling data
bem/watershed/BEM surface segmentation data (watershed algorithm)
bem/inner_skull.surfInner skull surface for BEM
bem/outer_skull.surfOuter skull surface for BEM
bem/outer_skin.surfSkin surface for BEM
sample-head.fifSkin surface in FIFF format for visualizations
surf/Surface reconstructions
mri/T1/The T1-weighted MRI data for visualizations

A second subject (morph) is provided for demonstrating surface morphing capabilities.

Preprocessing Already Completed​

The following steps have been already accomplished in the sample dataset:

  1. MRI surface reconstructions computed using FreeSurfer
  2. BEM surfaces created with the watershed algorithm
  3. MEG/EEG raw data checked and corrected
  4. Template scripts for averaging and noise-covariance written

Step-by-Step Walkthrough​

1. Setting Up​

# Set environment variables
export SUBJECTS_DIR=<yourdir>/subjects
export SUBJECT=sample

2. Setting Up Subject-Specific Data​

Structural MRIs:

mne_setup_mri

Source space (octahedron subdivision 6 ≈ 4098 sources/hemisphere):

mne_make_source_space --ico -6

Boundary-element models:

# Single-layer BEM (MEG only, faster)
mne_setup_forward_model --homog --surf --ico 4

# Three-layer BEM (MEG + EEG)
mne_setup_forward_model --surf --ico 4

3. Previewing the Data​

Before averaging, preview the raw data:

  1. Examine all channels and identify noisy or flat channels
  2. Mark bad channels using mne_mark_bad_channels or interactively
  3. Verify that the bad channels are correctly excluded
tip

You should be able to identify two problematic channels among the MEG and EEG channels in this dataset — one flat and one noisy.

4. Off-Line Averaging​

Compute averaged evoked responses using the averaging script:

mne_process_raw --raw sample_audvis_raw.fif \
    --lowpass 40 --projoff \
    --saveavetag -ave --ave audvis.ave

This creates sample_audvis-ave.fif containing averages for left and right auditory as well as left and right visual field stimuli.

5. Inspecting the Averages​

Load the averaged data in MNE Analyze or a compatible viewer and observe:

Auditory responses:

  • The main deflection occurs around 100 ms over the left and right temporal areas
  • The left-ear response is stronger on the right hemisphere; the opposite for the right-ear response

Visual responses:

  • Left and right visual field responses have quite different spatial distributions in the occipital area
  • A later response in the right parietal area is almost identical for both visual stimuli

6. Computing the Noise-Covariance Matrix​

mne_process_raw --raw sample_audvis_raw.fif \
    --lowpass 40 --projon \
    --savecovtag -cov --cov audvis.cov

This creates sample_audvis-cov.fif. The projections are set on, and the projection information is attached to the noise-covariance matrix.

7. MEG-MRI Coordinate Alignment​

Use MNE Analyze for the coordinate alignment. The process involves:

  1. Load the raw data file (provides digitizer data)
  2. Load an inflated surface for subject sample
  3. Identify the fiducial landmarks (nasion, LAP, RAP) on the MRI surface
  4. Click Align using fiducials for an initial alignment
  5. Discard outlier digitizer points (> 10 mm from surface)
  6. Use ICP (Iterative Closest Point) alignment to refine (10–20 iterations)
  7. Save the coordinate transformation
warning

Use the ICP alignment with caution. The iteration will not converge to a reasonable solution unless an initial alignment is performed first using fiducials.

8. Computing the Forward Solution​

cd MEG/sample

mne_forward_solution \
    --src $SUBJECTS_DIR/sample/bem/sample-oct-6-src.fif \
    --bem $SUBJECTS_DIR/sample/bem/sample-5120-5120-5120-bem-sol.fif \
    --meas sample_audvis-ave.fif \
    --trans sample-trans.fif \
    --meg --eeg --mindist 5 \
    --fwd sample_audvis-ave-oct-6-fwd.fif

For MEG-only analysis with the single-compartment BEM:

mne_forward_solution \
    --src $SUBJECTS_DIR/sample/bem/sample-oct-6-src.fif \
    --bem $SUBJECTS_DIR/sample/bem/sample-5120-bem-sol.fif \
    --meas sample_audvis-ave.fif \
    --trans sample-trans.fif \
    --meg --mindist 5 \
    --fwd sample_audvis-ave-oct-6-meg-fwd.fif

9. Computing the Inverse Operator​

Combined MEG+EEG inverse operator with depth weighting and loose orientation constraint:

mne_inverse_operator \
    --fwd sample_audvis-ave-oct-6-fwd.fif \
    --noisecov sample_audvis-cov.fif \
    --depth --loose 0.2 --meg --eeg

Separate MEG-only and EEG-only inverse operators:

# MEG only
mne_inverse_operator \
    --fwd sample_audvis-ave-oct-6-fwd.fif \
    --noisecov sample_audvis-cov.fif \
    --depth --loose 0.2 --meg

# EEG only
mne_inverse_operator \
    --fwd sample_audvis-ave-oct-6-fwd.fif \
    --noisecov sample_audvis-cov.fif \
    --depth --loose 0.2 --eeg
note

If you computed the forward solution with a single-compartment BEM, you can only compute the MEG inverse operator.

10. Interactive Analysis​

Explore the data and current estimates in MNE Analyze or MNE Inspect:

  1. Load surfaces — inflated surface for subject sample
  2. Load data — select sample_audvis-ave.fif and the inverse operator
  3. View field maps — observe magnetic field and potential maps at the N100m peak
  4. View current estimates — investigate the effects of regularization and other parameter settings
  5. Create labels — define ROIs at the auditory cortices and examine time courses
  6. Try morphing — load the morph subject and switch between original and morphed surfaces

See Also​